In operando study of orthorhombic V 2 O 5 as positive electrode materials for K-ion batteries Journal of Energy Chemistry

Herein, the electrochemical performance and the mechanism of potassium insertion/deinsertion in orthorhombic V 2 O 5 nanoparticles are studied. The V 2 O 5 electrode displays an initial potassiation/de-potassiation capacity of 200 mAh g (cid:1) 1 /217 mAh g (cid:1) 1 in the voltage range 1.5–4.0 V vs. K + /K at C/12 rate, suggesting fast kinetics for potassium insertion/deinsertion. However, the capacity quickly fades during cycling, reaching 54 mAh g (cid:1) 1 at the 31st cycle. Afterwards, the capacity slowly increases up to 80 mAh g (cid:1) 1 at the 200th cycle. The storage mechanism upon K ions insertion into V 2 O 5 is elucidated. In operando synchrotron diffraction reveals that V 2 O 5 ﬁrst undergoes a solid solution to form K 0.6 V 2 O 5 phase and then, upon further K ions insertion, it reveals coexistence of a solid solution and a two- phase reaction. During K ions deinsertion, the coexistence of solid solution and the two-phase reaction is identiﬁed together with an irreversible process. In operando XAS conﬁrms the reduction/oxidation of vanadium during the K insertion/extraction with some irreversible contributions. This is consistent with the results obtained from synchrotron diffraction, ex situ Raman, X-ray photoelectron spectroscopy (XPS), and transmission electron microscopy (TEM). Moreover, ex situ XPS conﬁrms the ‘‘cathode electrolyte interphase” (CEI) formation on the electrode and the decomposition of CEI ﬁlm during cycling. (cid:1) 2021 The Authors. Published by ELSEVIER B.V. and Science Press on behalf of Science Press and Dalian Institute of Chemical Physics, Chinese Academy of Sciences. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Along with their rapid development in modern society, renewable energy sources, such as solar, wind, geothermal power, and biomass energy, are playing crucial roles as power supplies. In particular, solar and wind powers gradually take an increasing share in energy supply. However, they are inherently intermittent and generally dispersed, strongly depending on the weather, time, and season, as well as location, while the consumption and demands of electric energy are comparatively constant [1]. Therefore, rechargeable batteries, being large-scale electrochemical energy storage systems (EESs), are necessary to realize the smooth integration of these intermittent energies into grids [2].
The first rechargeable potassium battery concept was designed by Eftekhari [3] in 2004, who employed Prussian blue (PB) positive electrode and potassium metal within 1 M KBF 4 in 3:7 of ethylene carbonate (EC):ethylmethyl carbonate (EMC) as electrolyte. Since then, potassium-ion batteries (KIBs) are becoming promising candidates to replace lithium-ion batteries (LIBs) for large-scale EESs. One of the main advantages is the abundance and low cost of potassium. Moreover, even if the K-ion is larger and heavier than Li + , Na + , and Mg 2+ , the K + /K couple has a standard electrode potential (À2.936 V vs. SHE) lower than that of Na + /Na (À2.714 V vs. SHE) and Mg 2+ /Mg (À2.37 V vs. SHE) [4]. Considering the gravimetric capacity, it is not appropriate just to compare the atomic weight of metals to calculate it. Instead, one should compare the formula weight of electrode materials as stated by Komaba et al. [5]. For example, the mass ratio of (P2-type K 2/3 CoO 2 ):(O3-LiCoO 2 ) is 1.19, giving a 19% increase of the positive electrode material mass despite that the K atom is 5.7 times heavier than Li (M K :M Li = 39/6. and K) does not significantly influence the total weight of the battery. In addition, K-ion in propylene carbonate (PC) displays a lower ion-solvent interaction due to a lower desolvation energy and a smaller solvated ion compared to Li-and Na-ion, resulting in fast diffusion kinetics and high rate capability [6]. Another advantage is that Al can be used as the current collector in anodes for both sodium-ion batteries (SIBs) and KIBs, whereas an expensive and heavier Cu foil must be used in anodes for LIBs, due to the Li-Al alloy formation at low potential [7,8]. Moreover, the fact that graphite electrodes can deliver a high reversible capacity of 279 mAh g À1 , with the formation of KC 8 [9], makes KIBs more advantageous than SIBs, where graphite cannot be used. Recently, other materials such as soft and hard carbon, Ti-based materials, transition metal sulfides, black P and organic compounds are also reported as negative electrodes for KIBs [10,11]. However, due to the large ionic radius of K + (1.38 Å), KIBs are still in their infancy and facing big challenges to develop electrode materials suitable for commercialization. To date, only limited positive electrode materials have been reported, including Prussian blue analogous, polyanionic, and organic compounds, as well as layered material (Co-, Mn-, Cr-, and V-based oxides) [10,11]. Among them, the Vbased oxides are very promising for the application in KIBs. For example, Clites and co-workers [12] recently reported that d-K x V 2 O 5 ÁnH 2 O positive electrode with an interlayer spacing of 9.65 Å shows good electrochemical performance towards Kinsertion/deinsertion with an initial potassiation capacity of 268 mAh g À1 at C/50 and of 226 mAh g À1 at C/15 rate in the potential range of 2.0-4.3 V vs. K + /K. The layered K 0.5 V 2 O 5 , investigated by Deng et al. [13], delivers a reversible capacity of 90 mAh g À1 at a current density of 10 mA g À1 and displays capacity retention of around 80% after 250 cycles at 100 mA g À1 in the potential range of 1.5-3.8 V vs. K + /K. Ex situ X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS) demonstrate that K 0.5 V 2 O 5 is highly stable, undergoing reversible structural changes during K-ions insertion and extraction. Zhu et al. [14] reported a bilayered d-K 0.51 V 2 O 5 positive electrode, which shows a capacity of 131 mAh g À1 with excellent rate capability and a high average voltage of 3.2 V. A V 2 O 5 Á0.6H 2 O xerogel positive electrode, which was reported by Tian et al. [15], delivers an initial potassiation capacity of 224 mAh g À1 at 50 mA g À1 with a value of 103 mAh g À1 after 500 cycles in the potential range 1.5-4.0 V vs. K + /K. Zhang et al. [16] reported a new layered compound orthorhombic K 0.83 V 2 O 5 with space group of Pnma, delivering an initial depotassiation/potassiation capacity of 86/90 mAh g À1 with a plateau of 3.5 V (vs. K + /K) and high capacity retention of 80% over 200 cycles.
Orthorhombic layered V 2 O 5 materials are intensively studied as a model system for intercalaction, but only few works focus on a study of these materials in KIBs. For example, in a work related to the V 2 O 5 Á0.6H 2 O xerogel, Tian et al. [15] reported on the performance of crystalline V 2 O 5 . However, this material displays a potassiation insertion capacity of only 44.3 mAh g À1 at 50 mA g À1 . Very recently, Ye and co-authors [17] reported on an amorphous V 2 O 5 /carbon nanotube sponge with an initial potassiation capacity of 206 mAh g À1 at 5 mA g À1 . However, the observed capacities are much lower than the theoretical capacity of 295 mAh g À1 (calculated considering 2 K-ions insertion) and no detailed K-storage mechanism of V 2 O 5 is proposed up to now.
In operando synchrotron powder diffraction and in operando Xray absorption spectroscopy (XAS) are very powerful tools to uncover the evolution of long-and short-range crystal structures and oxidation states of the transition metals during the electrochemical reactions [18]. The extremely bright, high flux, and high-energy synchrotron radiation provides many advantages for in operando studies in battery materials, for example, (1) the easy penetration through the sample, (2) high-quality data, (3) providing a ''real-time study" of time-dependent reactions.
In operando synchrotron diffraction and in operando XAS have not yet been employed to look into the detailed storage mechanism of a-V 2 O 5 in KIBs. As demonstrated by first-principles calculations, the migration barrier for K + in a monolayer of V 2 O 5 is reduced to 0.39 eV, which is much lower than in bulk V 2 O 5 (1.66 eV) [19,20]. Therefore, in our work, nanostructured V 2 O 5 materials are prepared to potentially enhance the electrochemical activity of V 2 O 5 positive electrode and to investigate the reaction mechanism during cycling. The present work focuses on the investigation of the electrochemical properties of V 2 O 5 nanoparticles in 1 M KPF 6 /PC and its storage mechanism during K ions insertion and extraction by in operando synchrotron diffraction and in operando XAS.

Synthesis of V 2 O 5 nanoparticles
The V 2 O 5 nanoparticles were synthesized through a hydrothermal method. Typically, 1.0 mL of 2 M HCl, 0.3 g of ammonium metavanadate (NH 4 VO 3 ), and 0.5 g of surfactant P123 (EO 20 PO 70 -EO 20 , where EO and PO represent ethylene oxide and propylene oxide, respectively) purchased from Sigma Aldrich were mixed into 30 mL of deionized water under ultrasonication for 10 min and then under stirring for 1 h. After that, the mixed solution was transferred to 50 mL Teflon-lined autoclaves and maintained at 120°C for 24 h with intermittent stirring in an oven. The resulting precipitate was filtered and washed with water and acetone several times, then dried under vacuum at 120°C for 24 h. The product was annealed at 400°C for 2 h in air after heating with a rate of 10°C/min.

Preparation of electrolyte
The preparation of the electrolyte was carried out in a glovebox (MBraun) under Ar atmosphere with very low H 2 O and O 2 content (<2 ppm). 1 M KPF 6 /PC electrolyte was prepared by dissolving a corresponding amount of KPF 6 powder in PC under vigorous stirring at room temperature in the glovebox. All chemicals are battery grade and purchased from Sigma-Aldrich.

Morphological and structural study
The morphology was studied with a Zeiss Supra 55 scanning electron microscope (SEM) with primary energy of 15 keV. The structural characterizations were done using synchrotron radiation (k = 0.4131 Å, 30 keV) at the Material Science and Powder Diffraction beamline (MSPD) at the ALBA synchrotron source (Barcelona, Spain) [21]. The powder diffraction patterns were measured in capillary geometry with powders filled in 0.5 mm Ø boro-silicate capillaries. A LabRam HR Evolution Raman microscope from Horiba Scientific equipped with HeNe laser (633 nm, 17 mW) and a CCD detector (Horiba) was used to collect the Raman scattering of the samples. Meanwhile, a 600 grating was used to split the measurement signal with a x 100 objective (NA 0.95) for all the pristine and cycled samples. The data was collected for 30 s with 4.25 mV of the laser and five measurements were added to reduce signal noise. In addition, the potassiated and depotassiated electrodes were sealed in an in situ Raman cell with a quartz window inside a glovebox. For transmission electron microscopy (TEM), all samples were pre-pared under an argon atmosphere inside a glovebox. The samples dispersed onto a carbon-coated grid by dry powder deposition were transferred to the TEM by using a Gatan TEM vacuum transfer holder. The TEM samples were examined using a Titan 80-300 electron microscope (FEI), equipped with a CEOS image spherical aberration corrector, high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) detector and a Tridiem Gatan image filter (GIF). The microscope was operated at an accelerating voltage of 300 kV. X-ray photoelectron spectroscopy measurements were performed using a K-Alpha + XPS spectrometer (ThermoFisher Scientific, East Grinstead, UK), applying a microfocused, monochromated Al K a X-ray source with a spot size of 400 mm. The sample storage and transportation were done in an airtight transport vessel under Ar to the spectrometer. To prevent any localized charge buildup, the K-Alpha + charge compensation system was employed during analysis, using electrons of 8 eV energy and low-energy argon ions. The Thermo Avantage software was used for data acquisition and processing [22]. The spectra were fitted with one or more Voigt profiles (binding energy uncertainty: ±0.2 eV). All spectra were referenced to the O 1s peak of vanadium oxide at 530.0 eV binding energy controlled by means of the well-known photoelectron peaks of metallic Cu, Ag, and Au.

Electrochemical characterization
The electrode slurry/mixture was prepared by mixing the active material (consisting of V 2 O 5 nanoparticles) with Super C65 (TIM-CAL) and polyvinylidene difluoride (PVDF) binder in a weight ratio of 70:20:10 with N-methyl-2-pyrrolidone solvent. The slurry was coated on 15 mm thick Al foil and dried at 70°C overnight. The electrodes (1 mg cm À2 ) were cut into discs with 12 mm diameter and then dried in the vacuum oven at 110°C for 12 h. Electrochemical performance was studied in 2032-type coin cells using potassium as the negative and V 2 O 5 as the positive electrode with Whatman separator in 1 M KPF 6 /PC electrolyte. The cells were assembled in a glovebox filled with Ar. Galvanostatic cycling experiments were conducted with a rate of C/12 on a multichannel electrochemical workstation (Bio-Logic VMP). For ex situ Raman and XPS, the cells were disassembled after cycling and the electrodes washed with PC in an Ar-filled glovebox. These samples consist of a V 2 O 5 electrode potassiated one time at a lower cut-off voltage of 1.5 V and a sample first potassiated and then depotassiated one time at an upper cut-off voltage of 4.0 V with C/12 rate.

In operando synchrotron diffraction and in operando XAS
In operando synchrotron diffraction was performed at the Material Science and Powder Diffraction beamline (MSPD) at the ALBA synchrotron. The in operando cells were built using 2025-type coin cells with glass windows of 5 mm diameter beam entrance. The positive electrode (active materials, 2 mg, and~7 mg cm À2 ) was prepared by pressing the electrode mixture on Al mesh within a 5 mm hole in the center, and K negative electrode with a 5 mm hole in the center. More details of this setup can be found elsewhere [23]. In operando synchrotron diffraction data were collected at room temperature with radiation k = 0.4131 Å wavelength (30 keV) and position sensitive detector MYTHEN. Data were collected with an effective exposure time of 60 s in steps of 0.006°over the angular range of 1.8°-42°in 2theta during the first cycles with C/15 rate (cell was running at a rate of C/30 for the first 5.4 h). The coin cells were continuously oscillated ±5°a round the incoming beam direction to improve the powder averaging (increasing the number of crystallites contribution). The diffraction data analysis was carried out by the Rietveld method using the Fullprof software package. In operando XAS measurements were performed at the synchrotron beamline P65 at PETRA III (DESY, Hamburg). Electrochemical cycling was conducted in 2025-type coin cell with 5 mm-diameter Kapton window for the first potassiated-depotassiated processes with C/14 rate. XAS spectra at the V K-edge were recorded in quick-XAS (6 min/spectrum) mode in fluorescence geometry using a PIPS (passivated implanted planar silicon) diode. The V K-edge for V 2 O 5 was measured for the 1st cycle and the energy was calibrated utilizing a vanadium foil as commonly applied in XAS experiments. V 2 O 3 , VO 2 , and V 2 O 5 were used as reference materials for V 3+ , V 4+ and V 5+ , respectively. All the data were collected at room temperature with a Si(1 1 1) double crystal monochromator and all the XAS spectra were processed using the DEMETER software package.  (Table S1), consistent with reported work [24]. Note that the common space group for pristine a-V 2 O 5 is Pmmn [25]. Actually, both space groups Pmn2 1 and Pmmn belong to the orthorhombic system. Pmn2 1 is acentric, while Pmmn is centrosymmetric because of the existence of additional inversion symmetry in Pmmn. Rietveld refinement based on both models demonstrated that a smaller Chi and R ffactor can be observed when acentric space group is used (Fig. S1, Tables S1-S3), suggesting it is more reasonable when acentric Pmn2 1 is used. It is impossible to distinguish the two space groups via X-ray powder diffraction analysis and we prefer to use Pmn2 1 because of the sigle crystal results by Novák et al. [24]. Moreover, b-axes is perpendicular to the [VO 5 ] square pyramids layer of V 2 O 5 in space group Pmn2 1 , while c-axes is perpendicular to that of V 2 O 5 in space group Pmmn (Fig. S1). Rietveld refinement shows the average apparent size of 62 nm with a standard deviation of 33 nm. Scanning electron microscopy (SEM) in Fig. 1(b) confirms that the V 2 O 5 consists of nanoparticles (around 50-200 nm), in agreement with Rietveld refinement result. During the first potassiation process, a short slope is observed until around 1.8 V, followed by a sloped plateau at about 1.7 V. In contrast, only one slope appears up to 4.0 V in the first depotassiation process. During the second cycle, only sloped curves can be observed, while a pair plateau at 2.6 V/2.85 V can be seen in the 5th cycle for the potassiation/depotassiation processes, respectively. V 2 O 5 nanoparticles deliver an initial potassiation/depotassiation capacity of 200 mAh g À1 /217 mAh g À1 . While the potassiation corresponds to the insertion of 1.36 K + per formula unit, the increased depotassiation capacity might be due to the decomposition of the electrolyte with the formation of the ''cathode electrolyte interphase" (CEI) (Fig. S2) and some other irreversible side reactions, where CEI terminology is widely used for the positive electrode side now [26][27][28]. In the subsequent potassiation (second cycle) the material delivers only 120 mAh g À1 and undergoes a continuous capacity loss down with only 94 mAh g À1 at the 5th cycle. The capacity reaches its lowest value of 54 mAh g À1 at the 31st cycle and then slowly increases up to 80 mAh g À1 at the 200th cycle, as shown in Fig. 2(b). The main reason for the decay could be due to the (partial) dissolution of V 2 O 5 , less formation of CEI (see XPS), the irreversibility of the materials, and side reaction during the cycling. In fact, the coulombic efficiency is low, which suggests the presence of parasitic side reactions. Understanding the cause of capacity increase with cycling is rather complicated and challenging since it depends on many factors and interplay of components present in the cell. We have tried to improve the cycling performance of V 2 O 5 by adding an additive fluoroethylene carbonate (FEC) to the electrolyte. V 2 O 5 only shows a limited initial potassiation/depotassiation capacity of 110/51 mAh g À1 with a large initial irreversible capacity. In the presence of FEC additive, the V 2 O 5 electrode shows better cycling stability and improved coulombic efficiency (close to 100% of coulombic efficiency after 5 cycles) but rather low and unsatisfying reversible capacity during 60 cycles compared with that in KPF 6 /PC as shown in Fig. S3. The stabilization can be due to a suppression of side reactions, which can arise from the interaction between the electrolyte and the highly reactive K metal which is used as coun-ter and reference electrode. A similar stabilization effect by FEC is also observed with Na metal in NIBs [29]. In our group, Pfeifer et al. [30] demonstrated that metallic sodium can strongly react with organic carbonate-based electrolytes in Na-ion batteries and strongly affect their electrochemical performance of the investigated electrode material. In the same line, in order to understand if some kind of decomposition reactions occur in presence of metallic K, a preliminary experiment was done by adding K metal to both KPF 6 /PC and KPF 6 /PC + FEC. A massive color change is observed when K is immersed in KPF 6 /PC along with the storage time (Fig. S4). The K metal changes to slight purple after 2 min and dark-earth purple (2 h) and finally to black-purple (21 h) in KPF 6 /PC, indicating strong reactions occurring between K metal and KPF 6 /PC electrolyte, which can cause strong side reaction for V 2 O 5 electrode during cycling. The clear solution of KPF 6 /PC turns to yellow after 21 h, while the solution of KPF 6 /PC + FEC does not show many changes. This experiment indicates that FEC could form a protective layer on the surface of metallic K (in a similar way as for Na). The interphase of K metal and the composition of electrolytes can play a very important role on the performance of batteries. However, detailed investigations are out of the scope of this work.

Electrochemical mechanism
In order to elaborate the electrochemical storage mechanism upon K ions insertion and extraction in V 2 O 5 , in operando synchrotron diffraction was performed during the potassiation/depotassiation processes as shown in Fig. 3. Before potassiation, all reflections, except those marked with asterisk related to the Al cur-  10.85°(020), 11.04°(120), 12.37°(600), and 13.51°(610), stay at the same position and their intensities also slowly reduce in Region I. During the solid solution reaction in Region I, the lattice parameter a shows minor changes (see Fig. S6), while lattice parameters b and c exhibit a significant increase. This is attributed to the insertion of large radii of K + in a-V 2 O 5 host, extending the interlayer distance (b direction) and c direction. The a axis is parallel to the zigzag chains of connershared VO 5 pyramids and the structure cannot be strongly compressed or expanded in the direction a axis due to the hard frame of zigzag chains along the direction of a axis. Along with further K ions insertion (1.65 V, Region II), the shifted reflections obtained in Region I continuously shift to lower angles/stay at the same positions, while three new broad reflections, 6.78°, 7.75°, and 12.86°( pointed out by grey arrows, belong to new phase Ph1), appear and their intensities slowly increase until the end of the first potassiation at 1.5 V, indicating coexistence of a two-phase transition and a solid solution. Among them, the reflection at 7.75°emerges after insertion of 0.6 mol of K + per mol V 2 O 5 (89 mAh g À1 ). At the end of potassiation at 1.5 V (189 mAh g À1 ), most of the reflections from K 0.6 V 2 O 5 disappear and only some broad reflections (K x -V 2 O 5 , x = 1.28) can be observed (see the diffraction pattern of the potassiated state at 1.5 V in Fig. S7, 69th pattern). This might be attributed to the dissolution of V 2 O 5 into the electrolyte or the formation of a K x V 2 O 5 compound with low crystallinity [31]. The dissolution of V 2 O 5 into the electrolyte is confirmed by the color changes of separator above and inductively coupled plasma optical emission spectrometry (ICP-OES) result (Table S4). In addition, the two-phase transition reaction is not completed at the end of potassiation at 1.5 V. Le Bail refinement demonstrates that the reflections at the fully potassiated state consist of K 0.6 V 2 O 5 and a new phase Ph1 with lattice parameters a = 11.107 Å, b = 8.940 Å, and c = 3.233 Å, in space group Pmn2 1 (Fig. S8, K x V 2 O 5 is the average formula of two phases K 0.6 V 2 O 5 and a new phase Ph1). During the 1st depotassiation, three different features can be observed. Two broad new reflections at 4.13°and 6.11°emerge (pointed out by black arrows in Fig. 3) and their intensities increase, while some reflections, such as 6.79°, 6.90°, 7.73°, 8.04°, 9.03°, 12.85°, 13.56°, and 15.01°, shift to higher angles and slightly grow up. Additionally, other small broad peaks almost keep unchanged. It indicates that, upon K ions extraction from the structure, the K x V 2 -O 5 material undergoes a two-phase transition process accompanied with a solid solution. Le Bail refinement demonstrates that the reflections at the depotassiated state (4.0 V, K y V 2 O 5 ) are related to K 0.6Àd V 2 O 5 and a phase Ph2 with lattice parameters a = 11.128 Å, b = 8.761 Å, and c = 3.226 Å, in space group Pmn2 1 (Fig. S9, K y V 2 O 5 is the average formula of two phases K 0.6Àd V 2 O 5 and Ph2). The phase evolution of a-V 2 O 5 during the 1st cycle is summarized as Scheme 1. Note that the lattice parameters of new phase Ph1 and Ph2 are similar to that of V 2 O 5 with a doubling of the b-axis. Unfortunately, Ph1 and Ph2 cannot relate to any K-V 2 O 5 based known phases and it is not possible to determine the structure neither of the new phases formed Ph1 at the end of potassiation nor of Ph2 at the end of the depotassiation. The reflections do not return to their initial positions of pristine V 2 O 5 , indicating irreversibility during the first K insertion/extraction (see Fig. S7). V 2 O 5 displays a different reaction mechanism upon K-ions insertion from what is known for Li-ions insertion, where a-, e-, d-, c-, and x-phase can be observed depending on the amount of inserted Li-ions (x) [32]. Particularly, the a-phase (x < 0.1) and e-phase  into c-phase is observed for x > 1, and the c-phase can be reversibly cycled in the range of 0 x 2.0 while maintaining the c-type structure. The irreversible formation of x-phase with a rocksalttype structure emerges upon the third lithium insertion into the V 2 O 5 structure. Note that the particle size of V 2 O 5 in our work is in a wide range of 50-200 nm, which would improve the capacity of materials because of its specific area and short K-ion migration. The reaction mechanism could change when the materials are changing from bulk size to nano-size. Particle size and charge/discharge rates will strongly affect the capacity of materials and the amount of inserted K ions and therefore have a significant influence on the observed synchrotron diffraction result. For example, nanostructuring of MoS 2 suppresses the 1T-triclinic phase transition and shortens Li-ion diffusion path lengths, allowing MoS 2 nanocrystal assemblies to behave as nearly ideal pseudocapacitors [33].
In operando XAS was performed on the V 2 O 5 positive electrode material to probe the electronic and structural environments of Vions during the electrochemical process. As displayed in Fig. 5, the edge position of the V K-edge spectrum for the initial state is slightly lower than that of the reference V 2 O 5 spectrum, where V 2 O 5 , VO 2 , and V 2 O 3 are referred as standard materials with +5, +4, and +3 oxidation states, respectively. Hence, it can be confirmed that the oxidation state of V is mainly +5 in the initial state, in good agreement with XPS results discussed later. Moreover, an intense pre-edge peak on XANES data is observed for the V Kedge of pristine V 2 O 5 , which is due to the transitions from 1s to bound p-hybridized d-states [34,35] with the loss of centrosymmetry of the occupied V site. Indeed, the orthorhombic crystal structure of V 2 O 5 is composed of square pyramids [VO 5 ] layers that share corners and edges and the V-ions are five-fold coordinated by oxygens in a distorted tetragonal pyramid.
During the first potassiation, the position (energy) of the V Kedge continuously shifts to lower energy, suggesting the reduction of vanadium ions accompanied by the K + insertion into the V 2 O 5 structure. The energy of the V K-edge of the electrode potas-siated to 1.5 V (x = 0.91 in K x V 2 O 5 ) is almost in the middle between that of standard VO 2 and V 2 O 5 in good agreement with the electrochemical data (Fig. S10). One could also clearly see the shift from the first-derivative V K-edge XANES for the V 2 O 5 during cycling (Fig. S11), indicating the reduction of the vanadium. Meanwhile, the pre-edge peak (A in Fig. 5(a)) slowly shifts to lower energy and its intensity gradually decreases, implying the reduction of the vanadium and the changes of the local V environments during K + insertion, owing to the co-existence of distorted tetragonal pyramids and centrosymmetric VO 6 octahedra. The intensity and the position of the pre-edge feature depend on the amount of inserted K ions. Along with the K ions insertion, the intensity of the pre-edge peak decreases. The position of the pre-edge peak shows a redshift because of an increase in symme-Scheme 1. Phase changes of V 2 O 5 during the 1st cycle.  try and a decrease in the oxidation state. The increase in symmetry causes a decrease of the peak intensity due to the decreased probability of the 1s-3d transition [36]. The distorted tetragonal pyramids tend to change into less distorted VO 6 octahedra. This also lessens the Jahn-Teller distortion of the V-O octahedral units, which become more regular in shape [37]. Moreover, the V Kedge pre-edge centroid as a function of potential shows the energy changes along with cycling, in agreement with the edge shift above (Fig. S12), where similar plots have been shown by Banerjee et al. [38]. The edge resonance (B and B 0 in Fig. 5), which is related to the energy absorption by core electrons [35,39] does not show remarkable changes in both shape and intensity. The reduction of V 2 O 5 upon K insertion is quite different from the observations upon Mg insertion in the same material [40]. Upon Mg insertion, significant changes in both shape and intensity of the edge resonance can be observed [40]. Furthermore, only one distinct isosbestic point [41] (red arrows in Fig. 5(a)) is obtained at~5474 eV energy during the potassiation process, implying the occurrence of a two-phase reaction upon K ions insertion into the V 2 O 5 . Note that isosbestic point is commonly observed in XAS spectra and its presence implies that only two species contribute to the absorption around the isosbestic point with changing fractions of their concentration in the mixture [42][43][44]. This result is fully consistent with the conclusions from in operando synchrotron diffraction (Fig. 3).
During the first depotassiation, both the energy of the V K-edge and the pre-edge peak (A 0 in Fig. 5(b)) exhibit a reversible behavior and shift to higher energy, indicating the progressive oxidation of V in the material during depotassiation. The intensity of the pre-edge peak continuously increases during K-ions extraction from the material, indicating a symmetry reduction at the V ions site. Moreover, one distinct isosbestic point (red arrows in Fig. 5(b)) is also obtained at~5474 eV energy during the depotassiation process, indicating a two-phase reaction upon K-ions extraction from the structure. However, the energy of V K-edge and pre-edge peak at the potential of 4.0 V do not return to their initial energies but are slightly lower than that of pristine V 2 O 5 , which is consistent with in operando synchrotron diffraction and indicates an irreversible process.
To further study the short-range structure of the samples during K-ions insertion/extraction, Raman spectra were collected, as shown in Fig. 6 and Table [45,46]. After K-ions insertion (1st potassiation), significant changes can be observed and some peaks related to V 2 O 5 at 994, 528, and 284 cm À1 become weak and broad, while those at 701, 304, and 284 cm À1 completely disappear, while some broad and distinct new peaks appear at 226, 340, 455, 763, 880, and 931 cm À1 . Similarly, these significant changes in Raman spectra are also observed after Mg-ions insertion in V 2 O 5 [40]. Hence, these new peaks might be ascribed to the new phase of K x V 2 O 5 , where the broad peaks are probably attributed to the disordered local structure in the K x V 2 O 5 phase. It can be deduced that K ions are inserted into the interlayer of V 2 O 5 crystals and bonded with oxygen atoms ([KO 6 ] octahedra) in the V 2 O 5 structure to form the Krich K x V 2 O 5 phase. After K ions extraction (K y V 2 O 5 ), some new peaks, observed at 763, 880, and 931 cm À1 , still remain at the same position and several additional broad peaks can be observed between 200 and 700 cm À1 . This suggests that K x V 2 O 5 and K y V 2 O 5 have similar structure and the existence of a second phase in the depotassiated samples is confirmed. The appearance of peaks at 163 and 996 cm À1 indicates the existence of a layered structure similar to that of pristine V 2 O 5 . However, the material cannot return back to the initial state of V 2 O 5 after K-ions extraction. These results confirm that K-ions insertion leads to the formation of a new K x V 2 O 5 phase. This transformation is only partially reversible during the following extraction of K-ions in the first cycle as identified by in operando synchrotron diffraction and XAS.
The morphology and the elemental distribution of potassium inside the V 2 O 5 framework are studied by scanning transmission electron microscopy coupled with EELS spectrum imaging. Fig. 7 (a) displays a representative STEM-HAADF image of the pristine V 2 O 5 sample. The initial morphology of the sample includes an uneven nanoparticle shape as well as a wire-like one, both in the nano-size range (around 50-200 nm) and in good agreement with SEM result. The high-resolution TEM image presented in Fig. 7(b) shows a highly crystalline sample with an interplanar distance of 0.44 nm corresponding with the (010) plane of the orthorhombic V 2 O 5 . Fig. 7(c) provides the elemental maps of K and V based on the STEM-EELS spectrum imaging of the potassiated sample. It proves a uniform distribution of K inside the V 2 O 5 matrix. This result implies that some of the K-ions still remain in the structure after depotassiation (Fig. 7(d)) and, therefore, confirms the irreversibility of K-insertion into V 2 O 5 during cycling.
The surface chemistry and surface elemental composition of pristine V 2 O 5 , of the first potassiated V 2 O 5 , and of the first depotassiated V 2 O 5 were investigated using X-ray photoelectron spectroscopy. According to Fig. 8, the V 2p spectra of all three samples can be fitted with two doublets: one with V 2p 3/2 at 517.2 eV and the other one with weak intensity at 515.9 eV [47], which refer to the oxidation states of vanadium +5 and +4, respectively. This indicates that V exists mainly in the oxidation state +5 with an average oxidation state of +4.9. The O 1s spectra of all three V 2 O 5 samples can be fitted with 3 peaks at 533.0, 531.4, and 530.0 eV, which correspond to the C-O and COO/CO 3 , as well as V-O groups [48,49], respectively. In the pristine V 2 O 5 , the minor V(IV) could be assigned to the reduction of residual surfactant P123 during synthesis. The V 2p spectrum in the potassiated state could be fitted with 1.8 at% V(V) at 517.0 eV and 1.2 at% V(IV) at 515.9 eV with an oxidation state of +4.6 (about 55 at% carbon, 26 at% oxygen, 11 at% potassium, and 5 at% F), while the V 2p spectrum of the depotassiated state shows two V 2p 3/2 components at 517.2 with 3.5 at% V(V) and 515.8 eV with 0.5 at% V(IV), returning back to its initial oxidation state of +4.9. In short, some of the V (V) is reduced to V(IV) during the electrochemical potassiation and then, partially oxidized upon electrochemical depotassiation, which is consistent with the above Raman, TEM, and in operando results. Furthermore, COO/CO 3 and C-O detected in the pristine V 2 O 5 could be due to the residual surfactant symmetric triblock copolymer P123 used during synthesis, where FTIR was performed to further confirm their existence (Fig. S13). After potassiation, the COO/CO 3 :V-O ratio significantly increases compared to that of pristine V 2 O 5 , indicating the formation of a CEI on the positive electrode particles. Upon depotassiation, the COO/CO 3 :V-O ratio decreases, but cannot return to the initial state, identifying the decomposition or destruction of the CEI film on the particles. Therefore, some of the recorded capacity is due to the film formation/decomposition and the other part is due to potassium insertion/deinsertion.

Conclusions
Orthorhombic a-V 2 O 5 nanoparticles were prepared via a hydrothermal approach. In the 1 M KPF 6 /PC electrolyte, V 2 O 5 nanoparticles deliver an initial potassiation/depotassiation capacity up to 200 mAh g À1 (1.36 K + )/217 mAh g À1 in the potential range of 1.5-4.0 V with a rate of C/12. The electrode shows a rapid capacity loss and exhibits the lowest capacity of only 54 mAh g À1 at the 31st cycle. After that, the capacity slowly increases up to a value of 80 mAh g À1 at the 200th cycle. The storage mechanism upon K ions insertion into V 2 O 5 is elucidated by in operando synchrotron diffraction and in operando XAS together with ex situ Raman, XPS, and TEM. Nano-V 2 O 5 can act as a host for the electrochemical insertion and extraction of K-ions. The K-insertion/extraction mechanism in a-V 2 O 5 differs from the ones for Li-and Mginsertion/extraction because of the larger size of K ions. V 2 O 5 undergoes a solid solution to form K 0.6 V 2 O 5 phase and then, upon further K ions insertion, goes through coexistence of a solid solution and a two-phase reaction. The coexistence of a two-phase and a solid solution mechanism reflects significant deviations from equilibrium, probably due to kinetic barriers and transport limitations. In addition, significant side reactions are indicated by the formation and destruction of a CEI, which might play a peculiar role in the degradation mechanism. Several strategies can be used to depress capacity fading such as carbon coating of electrode materials, additive for electrolytes, and producing stable passivation film on K anode. This approach helps to understand the mechanism occurring in electrode materials and benefits the development of KIBs towards improved electrochemical performance.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.